For principal component analysis (PCA) method，the influence of features on classification and recognition is not considered in the process of features fusion, which leads to the abnormal and non-effective separation of features after dimensionality reduction. Therefore, a prescreening PCA method was proposed to extract the time domain features and frequency domain features of signals. The correlation coefficient method was used to distinguish the relationship between objects effectively. The features that were not conducive to classification were removed, and the PCA dimension reduction was performed on the newly obtained matrix to transform the timedomain features and frequencydomain multifeatures into comprehensive evaluation indicators, so as to obtain better classification effect. The results show that this method has better feature separation effect. The results are helpful to improve the accuracy of PCA classification.